Test scsi_debug timeout and abort handling

[bvanassche]

A pending Linux kernel patch modifies SCSI abort handling in the
scsi_debug kernel driver. Hence, add a test that triggers aborting SCSI
commands for the scsi_debug kernel driver.

[bvanassche]

@kawasaki Do you expect me to ping you if a pull request is ready for review?

[kawasaki]

@bvanassche Sorry for the slow response. I put review of this PR on my todo list, but other work took higher priority... This ping helps me to prioritize blktests works.

I guess the kernel side change is this one, right?
https://lore.kernel.org/linux-scsi/[email protected]/
I applied this series on top of v6.18-rc2 kernel and ran the new test case scsi/012. I expected the test case pass, but it failed:

scsi/012 (test scsi_debug timeout and abort handling)        [failed]
    runtime  4.897s  ...  4.096s
    --- tests/scsi/012.out      2025-10-24 09:31:03.665000000 +0900
    +++ /home/shin/Blktests/blktests/results/nodev/scsi/012.out.bad     2025-10-24 11:08:07.001000000 +0900
    @@ -1,2 +1,4 @@
     Running scsi/012
    +Unexpected: dd command succeeded
    +Unexpected: scsi_timeout() has not been called
     Test complete

If the failure is unique to my system, I will need to allocate my time to debug it.

[kawasaki]

This PR looks important and valuable, and the patches look good to me overall. I left just two nit comments.

[bvanassche]

I guess the kernel side change is this one, right? https://lore.kernel.org/linux-scsi/[email protected]/ I applied this series on top of v6.18-rc2 kernel and ran the new test case scsi/012. I expected the test case pass, but it failed:

It seems like there is a misunderstanding. My goal with this test is that it passes against all kernel versions. This means that this test should pass whether or not the patch series mentioned above has been applied.

I see that this test passes if I run it on top of a debug kernel and that this test fails if run on top of a release kernel. Let me take a closer look at this.

[bvanassche]

Tests scsi/012 and scsi/007 have been merged into scsi/007.

[kawasaki]

Thanks for updating the series. The changes look good to me. I have just started another round of confirmation run. Tomorrow, I will check the result then will merge this PR.